Estimated glucose disposal rate and its dual role in hyperlipidemia risk and mortality: a secondary analysis of retrospective cohort data from U.S. and Chinese

Abstract

Background

Hyperlipidemia is a major global public health issue and a significant risk factor for various chronic diseases, including cardiovascular disease and diabetes. Insulin resistance (IR) is closely associated with hyperlipidemia. Estimated glucose disposal rate (eGDR), a non-invasive tool for assessing IR, may have clinical utility in identifying hyperlipidemia and predicting its prognosis.

Methods

This study is a secondary analysis of retrospective cohort data based on publicly available databases, specifically the U.S. National Health and Nutrition Examination Survey (NHANES) and the China Health and Retirement Longitudinal Study (CHARLS)—incorporating both cross-sectional and longitudinal follow-up data to systematically evaluate the relationship between eGDR and the risk of hyperlipidemia and mortality. Multivariable weighted logistic regression models were employed to analyze the risk of hyperlipidemia, while Cox proportional hazards models were used to assess all-cause and cardiovascular disease (CVD) mortality. Generalized additive models and smooth curve fitting were applied to identify potential nonlinear relationships, and subgroup as well as sensitivity analyses were conducted to verify the robustness of the findings.

Results

In the NHANES cohort, each standard deviation increase in eGDR was associated with a 11.5% reduction in the risk of hyperlipidemia (OR = 0.885 [0.867, 0.903]), an 8.6% reduction in all-cause mortality (HR = 0.914 [0.892, 0.936]), and a 10.4% reduction in CVD mortality (HR = 0.896 [0.859, 0.936]). In the CHARLS cohort, each SD increase in eGDR was associated with a 7.1% reduction in the risk of hyperlipidemia (OR = 0.929 [0.905, 0.954]) and an 9.2% reduction in all-cause mortality (HR = 0.908 [0.869,0.949]). A nonlinear inverse relationship was observed between eGDR and the risk of hyperlipidemia, with evidence of a significant threshold effect. Kaplan–Meier survival curves demonstrated significantly lower all-cause and CVD mortality among individuals with higher eGDR levels. Stratified analyses indicated that eGDR showed strong consistency and predictive value across different population subgroups, with particularly pronounced effects observed among younger individuals and those with diabetes.

Conclusion

Our study demonstrates that eGDR, as an indicator of insulin sensitivity, is significantly associated with the risk of hyperlipidemia and mortality, including both all-cause and CVD mortality. Improving eGDR levels may help reduce the health burden associated with hyperlipidemia and supports its potential clinical application in hyperlipidemia management and metabolic disease risk assessment.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12944-025-02684-6.

Introduction

Hyperlipidemia is a common metabolic disorder characterized by abnormally elevated plasma lipid levels [1]. As a major risk factor for various chronic conditions—including cardiovascular disease (CVD) and diabetes—hyperlipidemia not only imposes a significant disease burden but is also associated with numerous adverse health outcomes, severely compromising patients’ quality of life and life expectancy [2, 3]. From 1999 to 2020, the age-adjusted mortality rate of hyperlipidemia-related cardiovascular disease in the United States rose from 36.33 to 99.77 per 1,000,000, with higher mortality observed among males, non-Hispanic populations, rural residents, and Black individuals [4]. Epidemiological studies indicate that with continued global socioeconom

ic development, the prevalence of hyperlipidemia is steadily increasing, presenting a major challenge to global public health [5].

Current treatment strategies primarily rely on lifestyle modifications and pharmacotherapy. However, individual responses to treatment vary considerably, and medications may induce adverse effects that reduce adherence and diminish therapeutic efficacy [6, 7]. Although multiple treatment options exist, there remains an urgent need to develop more effective intervention strategies and clearly defined therapeutic targets to optimize clinical outcomes in patients with hyperlipidemia.

The association between insulin resistance (IR) and hyperlipidemia has been widely documented. Hyperlipidemia is typically characterized by elevated plasma levels of triglycerides and low-density lipoprotein cholesterol (LDL-C), along with reduced high-density lipoprotein cholesterol (HDL-C). Studies have shown that IR contributes to the development of hyperlipidemia by disrupting the metabolism of triglycerides, HDL-C, LDL-C, and very-low-density lipoprotein cholesterol (VLDL-C) [8]. Saori et al. [9] reported that reduced insulin sensitivity is closely linked to altered expression of lipid metabolism-related genes in skeletal muscle, suggesting that disrupted lipid handling in muscle may play a critical role in IR-mediated dyslipidemia. Similarly, Mamatha et al. [10] emphasized that the molecular mechanisms underlying hyperlipidemia are strongly associated with IR and the broader spectrum of metabolic dysfunction.

Various techniques are available to assess IR. The hyperinsulinemic-euglycemic clamp (HIEC) is considered the gold standard but is invasive and labor-intensive, limiting its feasibility in large-scale epidemiological studies [11]. The homeostasis model assessment of insulin resistance (HOMA-IR), which estimates IR using fasting glucose and insulin levels, is more practical in static conditions but limited by the non-routine use of fasting insulin assays in the general population [12]. In contrast, estimated glucose disposal rate (eGDR) combines multiple clinical and metabolic parameters to provide a more comprehensive assessment of IR. eGDR has shown high concordance with HIEC measurements and superior performance in predicting metabolic complications related to insulin resistance [13, 14].

This study utilized data from two nationally representative cohorts—NHANES (U.S.) and CHARLS (China)—employing both cross-sectional and longitudinal analytical approaches to investigate the association between eGDR and hyperlipidemia, as well as its impact on all-cause and cardiovascular disease (CVD) mortality among individuals with hyperlipidemia. By thoroughly analyzing the clinical and metabolic correlates of eGDR in the context of hyperlipidemia, we aim to uncover its potential clinical utility and evaluate its predictive accuracy for mortality in affected populations. This research seeks to offer new insights into the role of eGDR in the management of hyperlipidemia and provide scientific evidence to inform public health policy and personalized treatment strategies.

Materials and methods

Study design and participants

This study was based on two independent nationally representative cohorts: NHANES and CHARLS. The National Health and Nutrition Examination Survey (NHANES), administered by the U.S. Centers for Disease Control and Prevention, employs a complex, multistage, stratified probability sampling design combining household interviews, mobile examination center assessments, and laboratory testing to collect health data from the U.S. population. All protocols were approved by the National Center for Health Statistics Ethics Review Board, and written informed consent was obtained from all participants. This study utilized data from nine NHANES cycles conducted between 2001 and 2018, encompassing a total of 91,351 individuals. Participants were excluded based on the following criteria: (1) age < 20 years (n = 41,150); (2) missing hyperlipidemia diagnosis information (n = 2255); (3) absence of eGDR data (n = 4627); (4) lack of follow-up outcome data (n = 74); and (5) missing key covariates (n = 9320). A total of 33,925 eligible participants were included in the final analysis.

The China Health and Retirement Longitudinal Study (CHARLS) is a nationally representative longitudinal survey conducted by the National School of Development at Peking University, targeting adults aged 45 years and older in mainland China. The study was approved by the Ethics Review Committee of the Peking University Health Science Center, and written informed consent was obtained from all participants. We used 2011 wave (Wave 1) data for the cross-sectional analysis. Mortality data from 2011 to 2020 were included for survival outcome assessment. Participants were excluded based on the following criteria: (1) age < 45 years (n = 648); (2) missing hyperlipidemia diagnosis data (n = 5767); (3) absence of eGDR information (n = 1869); (4) missing mortality or follow-up data (n = 714); and (5) missing key covariates (n = 262). A total of 8448 participants met the inclusion criteria and were included in the final analysis. The detailed flowchart of inclusion and exclusion criteria is presented in Fig. 1.

Fig. 1.

Flowchart of the study

Definition of eGDR and hyperlipidemia

Estimated glucose disposal rate (eGDR) was calculated using the following formula: eGDR (mg/kg/min) = 21.158 − (0.090 × WC) − (3.407 × HT) − (0.551 × HbA1c), where WC refers to waist circumference measured at the level of the umbilicus (in cm); HT is a binary variable indicating hypertension status (1 = diagnosed, 0 = not diagnosed); and HbA1c denotes glycated hemoglobin level, expressed as a percentage (%). Hyperlipidemia was defined as the presence of any one of the following five criteria: (1) total cholesterol (TC) ≥ 200 mg/dL; (2) triglycerides (TG) ≥ 150 mg/dL; (3) low-density lipoprotein cholesterol (LDL-C) ≥ 130 mg/dL; (4) high-density lipoprotein cholesterol (HDL-C) ≤ 50 mg/dL for females or ≤ 40 mg/dL for males; (5) current use of lipid-lowering medications.

Survival Outcomes

Baseline health data from NHANES were linked to the National Death Index (NDI) through the end of the 2019 calendar year to construct the longitudinal cohort for survival analysis. CVD-specific mortality was identified based on the International Classification of Diseases, 10th Revision (ICD-10), including codes I00–I09, I11, I13, I20–I51, and I60–I69.

In the CHARLS cohort, mortality status (alive/deceased) was obtained from Waves 2 (2013), 3 (2015), 4 (2018), and 5 (2020). Although all waves recorded the date of interview, only Waves 2 and 5 provided relatively precise information on the date of death. For participants with specific death dates, survival time was calculated from the baseline interview to the exact date of death. For those with unclear death dates, survival time was estimated using the median time between the baseline interview and the wave in which death was reported [15, 16].

Covariates

Covariates

To control for potential confounding, covariates were selected based on previous literature and characteristics of the study datasets [17, 18]. In the NHANES dataset, the following covariates were included: age, gender, race, education level, marital status, poverty income ratio (PIR), body mass index (BMI), estimated glomerular filtration rate (eGFR), smoking and drinking status, diabetes, liver disease, total energy intake, intake of protein, carbohydrates, and fats, and the Healthy Eating Index-2015 (HEI-2015) score. Similarly, the CHARLS dataset included: age, sex, education level, marital status, BMI, eGFR, smoking and drinking status, diabetes, and liver disease. Given differences in variable definitions between the two datasets, selected variables were harmonized to enhance comparability. Marital status was categorized as “married or cohabiting” versus “divorced, widowed, or never married.” Education level was classified as “high school or below” and “above high school.” Smoking and drinking behaviors were categorized as “never,” “former,” and “current,” based on self-reported status. Diabetes was defined as meeting any of the following criteria: self-reported physician diagnosis; HbA1c > 6.5%; fasting plasma glucose ≥ 7.0 mmol/L; random plasma glucose ≥ 11.1 mmol/L; 2-hour plasma glucose after oral glucose tolerance test ≥ 11.1 mmol/L; or current use of antidiabetic medications or insulin. Liver disease status was determined based on self-reported prior diagnosis. Dietary intake data were obtained from 24-hour dietary recall interviews. The Healthy Eating Index-2015 (HEI-2015) score, ranging from 0 to 100, was calculated based on the recall data, with higher scores indicating healthier dietary quality. Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation; details are provided in the Supplementary Material.

Statistical analysis

Statistical analysis

This study integrated and analyzed data from two large nationally representative health surveys: the U.S. National Health and Nutrition Examination Survey (NHANES) and China’s Health and Retirement Longitudinal Study (CHARLS). The NHANES dataset included nine independent survey cycles conducted between 2001 and 2018. Given the complex multistage sampling design, all statistical analyses were adjusted using appropriate sampling weights. Laboratory subsample weights were additionally corrected by a factor of 1/9 to account for the number of cycles. For CHARLS, we used 2011 wave (Wave 1) data for the cross-sectional analysis, while mortality follow-up extended through 2020.After normality testing, between-group differences in continuous variables were assessed using the Wilcoxon rank-sum test, while categorical variables were compared using the chi-square test. To evaluate the association between eGDR and hyperlipidemia, weighted logistic regression models were constructed. To examine the relationship between eGDR and mortality among individuals with hyperlipidemia, weighted Cox proportional hazards models were employed. Generalized additive models (GAMs) were used to generate smoothed trend curves and identify inflection points. Based on these results, segmented linear regression models were fitted using maximum likelihood estimation to identify the optimal threshold by testing values between the 5th and 95th percentiles of eGDR, selecting the value that maximized the likelihood. Stratified subgroup analyses were conducted to examine heterogeneity across population subgroups. Sensitivity analyses were performed by reanalyzing NHANES data without weighting and adjusting covariates strictly according to CHARLS specifications to further verify the robustness and consistency of the findings. Additionally, we assessed the independent contributions of the individual components of eGDR to mortality risk. Modified eGDR models were constructed by excluding hypertension, HbA1c, or waist circumference to evaluate the added value of the full composite eGDR index. All statistical analyses were conducted using R software (version 4.4.1) and EmpowerStats (version 4.2), with a two-sided p-value < 0.05 considered statistically significant.

Results

Participant Characteristics

A total of 33,925 NHANES participants were included in the final analysis, with a mean age of 49.523 ± 17.741 years; 50.559% were male (n = 17,152). Among them, 24,459 participants met the diagnostic criteria for hyperlipidemia. Compared to participants without hyperlipidemia, those with hyperlipidemia exhibited significant differences across multiple baseline characteristics: they were generally older, had a higher proportion of females, a greater percentage of White individuals, and lower levels of education (mostly below high school). They were also more likely to be married or cohabiting. Clinically, they had higher body mass index (BMI), lower estimated glomerular filtration rate (eGFR), and a greater proportion with a history of smoking and alcohol consumption. The prevalence of diabetes and liver disease was also significantly higher in this group. In terms of dietary intake, individuals with hyperlipidemia had lower levels of total energy, protein, carbohydrate, and fat intake compared to those without hyperlipidemia (Table 1).

Table 1.

Baseline characteristics of NHANES participants classified by hyperlipidemia

Variables	Total (n = 33925)	Non-hyperlipidemia(n = 9466)	Hyperlipidemia(n = 24459)	P-value
AGE, years	49.523 ± 17.741	41.983 ± 17.264	52.442 ± 17.050	< 0.0001
Gender				< 0.001
Male	17,152 (50.559%)	4982 (52.630%)	12,170 (49.757%)
Female	16,773 (49.441%)	4484 (47.370%)	12,289 (50.243%)
Race				< 0.001
Mexican American	5574 (16.430%)	1388 (14.663%)	4186 (17.114%)
Other Hispanic	2674 (7.882%)	678 (7.162%)	1996 (8.161%)
Non-Hispanic White	16,071 (47.372%)	4103 (43.345%)	11,968 (48.931%)
Non-Hispanic Black	6727 (19.829%)	2372 (25.058%)	4355 (17.805%)
Other Race	2879 (8.486%)	925 (9.772%)	1954 (7.989%)
Education level				< 0.001
Under high school	8120 (23.935%)	1864 (19.692%)	6256 (25.577%)
High school or above	25,805 (76.065%)	7602 (80.308%)	18,203 (74.423%)
Marital status				< 0.001
Married/with partner	20,670 (60.929%)	5300 (55.990%)	15,370 (62.840%)
Unmarried/divorced/widowed	13,255 (39.071%)	4166 (44.010%)	9089 (37.160%)
PIR	2.604 ± 1.626	2.611 ± 1.638	2.601 ± 1.621	0.588
BMI	28.991 ± 6.667	26.880 ± 6.466	29.808 ± 6.564	< 0.0001
eGFR, ml/min/1.73m2	93.291 ± 23.331	100.914 ± 21.923	90.341 ± 23.192	< 0.0001
Smoking status				< 0.001
Never	18,127 (53.433%)	5494 (58.039%)	12,633 (51.650%)
Former	8566 (25.250%)	1871 (19.765%)	6695 (27.372%)
Now	7232 (21.318%)	2101 (22.195%)	5131 (20.978%)
Drinking status				< 0.001
Never	4538 (13.377%)	1174 (12.402%)	3364 (13.754%)
Former	5774 (17.020%)	1139 (12.033%)	4635 (18.950%)
Now	23,613 (69.604%)	7153 (75.565%)	16,460 (67.296%)
Diabetes				< 0.001
Yes	8477 (24.987%)	1215 (12.835%)	7262 (29.691%)
No	25,448 (75.013%)	8251 (87.165%)	17,197 (70.309%)
Liver disease				0.003
Yes	1272 (3.749%)	308 (3.254%)	964 (3.941%)
No	32,653 (96.251%)	9158 (96.746%)	23,495 (96.059%)
Total energy, calories	2134.755 ± 1008.379	2270.281 ± 1078.315	2082.304 ± 974.954	< 0.0001
Protein, g	81.687 ± 43.219	86.050 ± 46.241	79.998 ± 41.870	< 0.0001
Carbohydrate, g	257.298 ± 127.899	270.954 ± 134.271	252.013 ± 124.949	< 0.0001
Total fat, g	81.189 ± 47.128	86.235 ± 50.714	79.236 ± 45.515	< 0.0001
HEI-2015	50.718 ± 13.589	50.599 ± 13.723	50.764 ± 13.536	0.317

PIR Family income-to-poverty ratio, BMI Body mass index, eGFR estimated glomerular filtration rate, HEI-2015 Healthy Eating Index 2015

For continuous variables, the mean and standard deviation are reported

For categorical variables, frequencies and percentages are provided

In the CHARLS cohort, 8,448 participants meeting inclusion criteria were analyzed, with a mean age of 59.616 ± 9.323 years and 46.709% being male (n = 3,946). Among them, 5,740 were diagnosed with hyperlipidemia. Compared with the control group without hyperlipidemia, patients with hyperlipidemia were younger and had a higher proportion of females. They had a higher average BMI, lower eGFR, and were more likely to be non-smokers and non-drinkers. The prevalence of diabetes was significantly higher in the hyperlipidemia group than in the control group (Table 2).

Table 2.

Baseline characteristics of CHARLS participants classified by hyperlipidemia

Variables	Total (n = 8448)	Non-hyperlipidemia(n = 2708)	Hyperlipidemia(n = 5740)	P-value
AGE, years	59.616 ± 9.323	60.099 ± 9.627	59.387 ± 9.168	0.001
Gender				< 0.001
Male	3946 (46.709%)	1614 (59.601%)	2332 (40.627%)
Female	4502 (53.291%)	1094 (40.399%)	3408 (59.373%)
Education level				0.254
Under high school	7658 (90.649%)	2469 (91.174%)	5189 (90.401%)
High school or above	790 (9.351%)	239 (8.826%)	551 (9.599%)
Marital status				0.671
Married/with partner	7406 (87.666%)	2368 (87.445%)	5038 (87.770%)
Unmarried/divorced/widowed	1042 (12.334%)	340 (12.555%)	702 (12.230%)
BMI	23.508 ± 3.936	22.205 ± 3.582	24.122 ± 3.946	< 0.001
eGFR, ml/min/1.73m2	91.640 ± 14.901	92.576 ± 14.718	91.199 ± 14.967	< 0.001
Smoking status				< 0.001
Never	5106 (60.440%)	1395 (51.514%)	3711 (64.652%)
Former	750 (8.878%)	236 (8.715%)	514 (8.955%)
Now	2592 (30.682%)	1077 (39.771%)	1515 (26.394%)
Drinking status				< 0.001
Never	4957 (58.677%)	1359 (50.185%)	3598 (62.683%)
Former	903 (10.689%)	342 (12.629%)	561 (9.774%)
Now	2588 (30.634%)	1007 (37.186%)	1581 (27.544%)
Diabetes				< 0.001
Yes	1259 (14.903%)	262 (9.675%)	997 (17.369%)
No	7189 (85.097%)	2446 (90.325%)	4743 (82.631%)
Liver disease				0.114
Yes	295 (3.492%)	107 (3.951%)	188 (3.275%)
No	8153 (96.508%)	2601 (96.049%)	5552 (96.725%)

BMI Body mass index, eGFR estimated glomerular filtration rate

For continuous variables, the mean and standard deviation are reported. For categorical variables, frequencies and percentages are provided

Association Between eGDR and Hyperlipidemia

In both the NHANES and CHARLS cohorts, eGDR was significantly and inversely associated with the prevalence of hyperlipidemia (Table 3). In NHANES, after adjustment for all potential confounders, each standard deviation (SD) increase in eGDR was associated with a 11.5% reduction in the odds of hyperlipidemia (OR = 0.885 [0.867, 0.903]). Quartile analyses demonstrated a clear dose–response relationship: compared with the lowest quartile (Q1), the odds of hyperlipidemia were educed by 17.2% in Q2 (OR = 0.828 [0.726, 0.944]), 17.1% in Q3 (OR = 0.829 [0.737, 0.933]), and 59.4%4 in Q4 (OR = 0.406 [0.350, 0.471]). Smoothed curve fitting indicated a nonlinear association between eGDR and hyperlipidemia (Fig. 2A). Segmented logistic regression and threshold effect analyses revealed that when eGDR < 9.514, hyperlipidemia risk decreased moderately with increasing eGDR (OR = 0.952 [0.935, 0.969]); however, for eGDR > 9.514, the risk dropped sharply (OR = 0.554 [0.527, 0.614]) (Supplementary Table 1). In the CHARLS cohort, each standard unit increase in eGDR was associated with a 7.1% reduction in hyperlipidemia risk (OR = 0.929 [0.905, 0.954]) after covariate adjustment. A similar dose–response trend was observed in the quartile analysis: compared to Q1, the odds of hyperlipidemia were reduced by 22.1% in Q2 (OR = 0.779 [0.673, 0.903]), 27,5% in Q3 (OR = 0.725 [0.625, 0.842]), and 42% in Q4 (OR = 0.580 [0.493, 0.682]). Figure 2B illustrates the nonlinear pattern of this relationship in CHARLS. Threshold effect analysis showed that when eGDR < 12.092, there was a significant inverse association with hyperlipidemia (OR = 0.912 [0.885, 0.939]), whereas this association was not statistically significant for eGDR > 12.092 (OR = 1.088 [0.971, 1.219]) (Supplementary Table 1).

Table 3.

The association between eGDR and hyperlipidemia

Characteristics	Model 1 OR (95% CI)	Model 2 OR (95% CI)	Model 3 OR (95% CI)
NHANES
Continuous eGDR	0.747 (0.735, 0.758) <0.001	0.794 (0.782, 0.807) <0.001	0.885 (0.867, 0.903) <0.001
eGDR quartiles
Q1	1[Ref]	1[Ref]	1[Ref]
Q2	0.571 (0.506, 0.643) <0.001	0.606 (0.536, 0.686) <0.001	0.828 (0.726, 0.944) <0.001
Q3	0.433 (0.393, 0.477) <0.001	0.558 (0.503, 0.620) <0.001	0.829 (0.737, 0.933) 0.002
Q4	0.515 (0.499, 0.532) <0.001	0.202 (0.182, 0.225) <0.001	0.406 (0.350, 0.471) <0.001
P for trend	< 0.001	< 0.001	< 0.001
CHARLS
Continuous eGDR	0.844 (0.825, 0.862) <0.001	0.838 (0.819, 0.857) <0.001	0.929 (0.905, 0.954) <0.001
eGDR quartiles
Q1	1[Ref]	1[Ref]	1[Ref]
Q2	0.682 (0.592, 0.785) <0.001	0.674 (0.584, 0.778) <0.001	0.779 (0.673, 0.903) <0.001
Q3	0.534 (0.465, 0.613) <0.001	0.522 (0.454, 0.602) <0.001	0.725 (0.625, 0.842) <0.001
Q4	0.321 (0.280, 0.367) <0.001	0.315 (0.274, 0.361) <0.001	0.580 (0.493, 0.682) <0.001
P for trend	< 0.001	< 0.001	< 0.001

OR Odd ratio, CI Confidence interval

For NHANES: Model 1: Unadjusted for covariates, Model 2: Adjusted for age, race, and gender, Model 3: Adjusted for age, gender, race, education level, marital status, PIR, BMI, eGFR, smoking and drinking status, diabetes, liver disease, total energy intake, protein intake, carbohydrate intake, and fat intake, and the Healthy Eating Index-2015

For CHARLS: Model 1: Unadjusted for covariates, Model 2: Adjusted for age and gender, Model 3: Adjusted for age, gender, education level, marital status, BMI, eGFR, smoking and drinking status, diabetes, and liver disease

Fig. 2.

Smooth curve fitting analysis of the association between eGDR and hyperlipidemia in NHANES(A) and CHARLS(B).For NHANES: Adjusted for age, gender, race, education level, marital status, PIR, BMI, eGFR, smoking and drinking status, diabetes, liver disease, total energy intake, protein intake, carbohydrate intake, and fat intake, and the Healthy Eating Index-2015. For CHARLS: Adjusted for age, gender, education level, marital status, BMI, eGFR, smoking and drinking status, diabetes, and liver disease

Figure 3 presents subgroup analyses of the association between eGDR (as a continuous variable) and hyperlipidemia risk, stratified by age, sex, BMI, smoking status, drinking status, diabetes, and liver disease. In NHANES, eGDR was significantly inversely associated with hyperlipidemia risk across all subgroups. Interaction tests suggested that this association was particularly pronounced among younger individuals (aged 20–44 years) and among those who had never smoked, suggesting higher susceptibility to reduced insulin sensitivity in these groups. In the CHARLS cohort, most subgroup analyses supported a significant inverse association between eGDR and hyperlipidemia risk. However, the association did not reach statistical significance among individuals with BMI > 29.9, those with a history of smoking or drinking, and those with liver disease—likely due to reduced statistical power from smaller subgroup sample sizes. Notably, no significant interactions were detected for any stratification variables in CHARLS, suggesting a robust and consistent relationship between eGDR and hyperlipidemia risk.

Fig. 3.

Subgroup analysis of the association between eGDR and hyperlipidemia in NHANES(A) and CHARLS(B).For NHANES: Adjusted for age, gender, race, education level, marital status, PIR, BMI, eGFR, smoking and drinking status, diabetes, liver disease, total energy intake, protein intake, carbohydrate intake, and fat intake, and the Healthy Eating Index-2015. For CHARLS: Adjusted for age, gender, education level, marital status, BMI, eGFR, smoking and drinking status, diabetes, and liver disease

Association Between eGDR and Mortality Among Individuals With Hyperlipidemia

During a mean follow-up of 9.33 years in the NHANES cohort, a total of 3,734 all-cause deaths (15.2%) and 1,208 cardiovascular disease (CVD) deaths (4.9%) were recorded among individuals with hyperlipidemia. When stratified by eGDR quartiles, all-cause mortality rates were 22.1% (n = 1,356) in Q1, 21.2% (n = 1,296) in Q2, 11.2% (n = 690) in Q3, and 6.4% (n = 392) in Q4. Corresponding CVD mortality rates were 7.9% (n = 484), 7.0% (n = 429), 3.3% (n = 205), and 1.4% (n = 90), respectively. Weighted Cox proportional hazards regression analysis revealed that eGDR was significantly inversely associated with both all-cause and CVD mortality. After adjustment for all potential confounders, each standard deviation increase in eGDR was associated with an 8.6% reduction in all-cause mortality (HR = 0.914 [0.892, 0.936]) and a 10.4% reduction in CVD mortality (HR = 0.896 [0.859, 0.936]). Quartile analysis supported a dose–response relationship (P for trend < 0.05), with individuals in the highest eGDR quartile (Q4) showing a 26% lower risk of all-cause mortality (HR = 0.740 [0.625, 0.875]) and a 38.3% lower risk of CVD mortality (HR = 0.617 [0.450, 0.845]) compared to Q1 (Table 4). Smoothed curve fitting demonstrated a linear inverse association between eGDR and both all-cause and CVD mortality, indicating that higher eGDR levels were closely associated with reduced mortality risk (Fig. 4).

Table 4.

Association between eGDR and all-cause mortality and CVD mortality in hyperlipidemic patients in the NHANES cohort

Characteristics	Model 1 h (95% CI)	Model 2 h (95% CI)	Model 3 h (95% CI)
All-cause mortality
Continuous eGDR	0.809 (0.798, 0.821) < 0.001	0.907 (0.890,0.923) < 0.001	0.914 (0.892,0.936) <0.001
eGDR index quartiles
Q1	1[Ref]	1[Ref]	1[Ref]
Q2	0.857 (0.773, 0.949) 0.003	0.853 (0.770, 0.945) 0.002	0.978 (0.869, 1.100) 0.709
Q3	0.383 (0.341, 0.429) < 0.001	0.686 (0.615, 0.765) < 0.001	0.799 (0.711, 0.898) < 0.001
Q4	0.211 (0.185, 0.240) < 0.001	0.594 (0.518, 0.680) < 0.001	0.740 (0.625, 0.875) < 0.001
P for trend	< 0.001	< 0.001	< 0.001
CVD mortality
Continuous eGDR	0.778 (0.761, 0.795) < 0.001	0.869 (0.842,0.897) < 0.001	0.896 (0.859,0.936) < 0.001
eGDR index quartiles
Q1	1[Ref]	1[Ref]	1[Ref]
Q2	0.755 (0.650, 0.877) < 0.001	0.750 (0.647, 0.870) < 0.001	0.931 (0.773, 1.122) 0.452
Q3	0.289 (0.236, 0.353) < 0.001	0.551 (0.674, 0.677) < 0.001	0.690 (0.548, 0.871) 0.001
Q4	0.132 (0.104, 0.168) < 0.001	0.437 (0.341, 0.559) < 0.001	0.617 (0.450, 0.845) 0.002
P for trend	< 0.001	< 0.001	< 0.001

HR Hazard ratio, CI Confidence interval

Model 1: Unadjusted for covariates, Model 2: Adjusted for age, race, and gender, Model 3: Adjusted for age, gender, race, education level, marital status, PIR, BMI, eGFR, smoking and drinking status, diabetes, liver disease, total energy intake, protein intake, carbohydrate intake, and fat intake, and the Healthy Eating Index-2015

Fig. 4.

Smooth curve fitting analysis of the association between eGDR and all-cause mortality (A) and CVD mortality (B) in patients with hyperlipidemia (NHANES).Adjusted for age, gender, race, education level, marital status, PIR, BMI, eGFR, smoking and drinking status, diabetes, liver disease, total energy intake, protein intake, carbohydrate intake, fat intake, and the Healthy Eating Index-2015

Kaplan–Meier survival curves stratified by eGDR quartiles are presented in Fig. 5. Both all-cause and CVD mortality showed progressively improved survival probabilities with increasing eGDR. The lowest quartile group (Q1) had the poorest survival, while the highest quartile (Q4) had the best outcomes. Log-rank tests indicated highly significant differences across quartile groups (p < 0.0001), suggesting that eGDR is a strong predictor of survival in hyperlipidemic individuals. Subgroup analyses further showed a consistent inverse association between eGDR and mortality risk across all demographic and clinical subgroups. Notably, a significant interaction effect was observed for age: the association between eGDR and both all-cause and CVD mortality was most pronounced in individuals aged 20–44 years, suggesting that younger populations may be more metabolically sensitive. Additionally, stratified analyses by diabetes status revealed stronger associations among individuals with diabetes, suggesting enhanced predictive value of eGDR in populations with insulin resistance (Fig. 6).

Fig. 5.

Kaplan–Meier survival analysis of hyperlipidemia patients in different eGDR quartile groups, all-cause mortality (A) and CVD mortality (B) (NHANES)

Fig. 6.

Subgroup analysis of the association between eGDR and all-cause mortality (A) and CVD mortality (B) in patients with hyperlipidemia (NHANES).Adjusted for age, gender, race, education level, marital status, PIR, BMI, eGFR, smoking and drinking status, diabetes, liver disease, total energy intake, protein intake, carbohydrate intake, fat intake, and the Healthy Eating Index-2015

In the CHARLS cohort, over a mean follow-up of 8.9 years, 540 all-cause deaths were documented among participants with hyperlipidemia, accounting for 9.4% of the cohort. When stratified by eGDR quartiles, all-cause deaths were 11.7% (n = 168) in Q1, 11.6% (n = 168) in Q2, 6.0% (n = 86) in Q3, and 8.2% (n = 118) in Q4. Cox regression analysis demonstrated a significant inverse association between eGDR and all-cause mortality (Table 5). After adjusting for all confounding factors, each standard deviation increase in eGDR was associated with an 9.2% reduction in mortality risk (HR = 0.908 [0.869,0.949]). Quartile analyses supported a dose–response trend (P for trend < 0.05), with participants in the fourth quartile (Q4) had a significantly lower risk of death than those in the first quartile (Q1) (HR = 0.649 [0.491,0.858]). Smoothed trend analysis (Figure S1) also indicated a linear inverse relationship between eGDR and all-cause mortality, further reinforcing its prognostic relevance.

Table 5.

Association between eGDR and all-cause mortality in hyperlipidemic patients in the CHARLS cohort

Characteristics	Model 1 h (95% CI)	Model 2 h (95% CI)	Model 3 h (95% CI)
All-cause mortality
Continuous eGDR	0.904 (0.872, 0.938) < 0.001	0.946 (0.911,0.982) 0.003	0.908 (0.869,0.949) <0.001
eGDR index quartiles
Q1	1[Ref]	1[Ref]	1[Ref]
Q2	0.998 (0.806,1.236) 0.987	1.079 (0.871, 1.336) 0.488	0.958 (0.767,1.198) 0.709
Q3	0.484 (0.373, 0.627) < 0.001	0.651 (0.502, 0.846) 0.001	0.593 (0.452, 0.777) <0.001
Q4	0.689 (0.544,0.872) 0.002	0.839 (0.662, 1.064) 0.147	0.649 (0.491,0.858) 0.002
P for trend	< 0.001	0.011	< 0.001

HR Hazard ratio, CI Confidence interval

Model 1: Unadjusted for covariates, Model 2: Adjusted for age and gender, Model 3: Adjusted for age, gender, education level, marital status, BMI, eGFR, smoking and drinking status, diabetes, and liver disease

Sensitivity Analysis

To validate the robustness of our findings, several sensitivity analyses were conducted. In unweighted models, the inverse association between eGDR and hyperlipidemia risk remained significant (Table S2). In fully adjusted models, each standard deviation (SD) increase in eGDR was associated with a 31.4% reduction in hyperlipidemia risk (OR = 0.686 [0.657, 0.715]); compared to the lowest quartile (Q1), the highest quartile (Q4) exhibited a 68.6% lower risk of hyperlipidemia (OR = 0.314 [0.257, 0.383]). Among individuals with hyperlipidemia, eGDR also remained significantly inversely associated with all-cause and CVD mortality. After full adjustment, each SD increase in eGDR was associated with a 15.5% reduction in all-cause mortality (HR = 0.845 [0.812, 0.880]) and a 15% reduction in CVD mortality (HR = 0.850 [0.791, 0.913]). This association was consistent across eGDR quartiles (P for trend > 0.05) (Table S3).

We further replicated the analysis by applying CHARLS-based covariate definitions to the NHANES dataset. The results remained robust and consistent (Table S4). In this fully adjusted model, each SD increase in eGDR was associated with an 11% reduction in hyperlipidemia risk (OR = 0.890 [0.872, 0.908]); participants in the highest quartile had a 58.1% lower risk than those in the lowest quartile (OR = 0.419 [0.362, 0.485]). Similarly, each SD increase in eGDR was associated with a 9% reduction in all-cause mortality (HR = 0.910 [0.887, 0.932]) and a 10.8% reduction in CVD mortality (HR = 0.892 [0.854, 0.930]). Compared with Q1, individuals in Q4 had a 28.9% lower risk of all-cause mortality (HR = 0.711 [0.601, 0.840]) and a 41.2% lower risk of CVD mortality (HR = 0.588 [0.433, 0.797]) (Table S5). To address the potential collinearity issues in the eGDR formula, we conducted a sensitivity analysis to evaluate the independent contribution of each component of eGDR to mortality risk. The results showed that when each component of eGDR was assessed individually, waist circumference, hypertension, and glycated hemoglobin all had a significant impact on mortality risk, and each component was independently associated with both all-cause mortality and CVD mortality (Table S6). Furthermore, when modified eGDR models were constructed by excluding waist circumference, hypertension, or glycated hemoglobin, the association between eGDR and mortality risk remained statistically significant, further supporting the robustness of the composite index (Table S7). As a composite measure, eGDR provided additional predictive value compared to the individual components, although each component itself still played an important role in predicting mortality risk.

Discussion

This study conducted a comprehensive analysis using cross-sectional and longitudinal data from the U.S. NHANES and China’s CHARLS cohorts to investigate the association between the insulin resistance (IR) indicator eGDR and the risk of hyperlipidemia, as well as all-cause and cardiovascular disease (CVD) mortality among individuals with hyperlipidemia. Our findings demonstrate that eGDR is significantly inversely associated with hyperlipidemia risk, and with both all-cause and CVD mortality in hyperlipidemic populations. Further analysis using smoothed curve fitting and segmented logistic regression revealed a threshold effect: eGDR values exceeding 9.514 were associated with a markedly lower risk of developing hyperlipidemia.

This study found that eGDR, an indicator of insulin resistance, is negatively correlated with the risk of hyperlipidemia in a nonlinear manner, validated across both U.S. and Chinese populations, supporting eGDR’s potential as a universal risk assessment tool. Previous studies support our conclusion, in states of IR, the regulatory role of insulin in lipid metabolism is diminished, leading to increased hormone-sensitive lipase activity and enhanced lipolysis in adipose tissue. This process releases large amounts of free fatty acids (FFA), which in turn promote hepatic very-low-density lipoprotein (VLDL) synthesis and elevate plasma triglyceride levels, thereby contributing to hyperlipidemia [19, 20]. IR also impairs hepatic clearance of LDL-C, further exacerbating its accumulation [21]. Conversely, hyperlipidemia can worsen IR by inhibiting insulin signaling pathways and GLUT4 translocation, thereby reducing glucose uptake and utilization in target tissues [22]. A study of 231 men with newly diagnosed hyperlipidemia found that plasma cis-vaccenic acid (cVA) levels were negatively correlated with IR indices, suggesting that cVA may modulate insulin sensitivity [23]. Another metabolomics-based study in obese individuals reported that short-term high-fat diets exacerbated insulin resistance and glucose-lipid dysregulation, further reinforcing the metabolic link between IR and hyperlipidemia [24].

Smoothed curves indicated a clear threshold effect. In NHANES, for eGDR < 9.514, increases in eGDR were associated with modest reductions in hyperlipidemia risk, suggesting a metabolic “plateau phase” where IR improvement may be offset by other metabolic abnormalities [25, 26]. Beyond this threshold, substantial reductions in risk were observed, likely due to regained insulin-regulated lipid metabolism—such as increased fatty acid oxidation, suppressed hepatic VLDL synthesis, and improved HDL-C function [27–29]. In contrast, CHARLS revealed that the protective effect of eGDR plateaued when exceeding 12.092, indicating a potential “diminishing returns” effect, where further improvements in insulin sensitivity confer limited additional benefit in metabolically healthy individuals [30, 31]. This difference may be attributed to the variations in age structure, genetic background, dietary patterns, and baseline metabolic health between the two populations. The NHANES cohort includes individuals aged 20 and older, while the CHARLS cohort includes individuals aged 45 and older, which could affect insulin sensitivity and metabolic health. Age-related insulin resistance and changes in lipid metabolism may contribute to a higher eGDR threshold in the CHARLS cohort. Furthermore, ethnic and genetic differences between populations in the U.S. and China may also influence the association between eGDR and dyslipidemia risk. Differences in dietary habits, such as variations in carbohydrate and fat intake, may also play a role in shaping the eGDR threshold. From a clinical perspective, interventions aimed at raising eGDR above the critical 9.514 threshold may yield substantial metabolic benefits. However, for individuals with already high eGDR, additional strategies targeting genetic predisposition, hepatic lipid metabolism, or other mechanisms may be warranted. Subgroup analyses further supported the robustness of eGDR’s association with hyperlipidemia risk. In NHANES, this inverse association was consistent across all subgroups, with particularly strong effects in younger adults (20–44 years) and never-smokers. These findings suggest a heightened “metabolic sensitivity” to insulin in individuals without significant comorbid burden or tobacco-induced oxidative stress [32–34]. Although similar trends were observed in CHARLS, statistical significance was not achieved in subgroups such as those with severe obesity, prior smoking or drinking, or liver disease, possibly due to limited sample size or more complex lipid dysregulation in these groups. Interestingly, in the NHANES baseline characteristics, the total energy, protein, carbohydrate, and fat intake in the dyslipidemia group were significantly lower than those in the non-dyslipidemia group [35]. This finding contradicts intuition, and the difference may be due to participants underestimating their dietary intake. Another explanation could be reverse causality, where individuals diagnosed with dyslipidemia may have adjusted their diet after the diagnosis. These factors should be considered when interpreting baseline data, and future studies using more objective dietary assessment methods will help clarify this issue.

Our study also highlights the prognostic utility of eGDR in predicting long-term mortality among individuals with hyperlipidemia. NHANES follow-up data over nearly a decade revealed that higher eGDR levels were significantly associated with reduced all-cause and CVD mortality. This finding is consistent with previous studies, In a study by Alemayehu et al. [36], 80.3% of 256 cardiac patients exhibited at least one lipid abnormality, with high TG (30.1%) and low HDL-C (72.5%) being most common. Karrowni et al. [37] demonstrated that in 1,073 non-diabetic acute myocardial infarction patients, higher HOMA-IR levels were associated with significantly increased prevalence of multi-vessel coronary artery disease, suggesting IR may drive atherosclerosis and elevate cardiovascular mortality risk. Similarly, in 241 patients with acute decompensated heart failure (ADHF), Nogi et al. [38] found that low serum insulin was an independent predictor of all-cause and CVD mortality over 21.8 months of follow-up. Kaplan–Meier survival curves showed clear improvements in survival probability with increasing eGDR quartiles, and log-rank tests confirmed statistically significant survival differences between groups. The association was more evident in younger persons and patients with diabetes, indicating that eGDR may have superior prognostic significance in metabolically active or insulin-resistant groups [14, 39]. These results underscore the need of early therapies to enhance insulin sensitivity, especially in high-risk populations, to attain substantial long-term survival advantages. Although CHARLS had fewer mortality events, results aligned with those from NHANES, underscoring the generalizability and external validity of eGDR as a prognostic biomarker.

This study has several strengths. First, it utilized two large, nationally representative datasets from the U.S. (NHANES) and China (CHARLS), allowing for cross-national validation of findings. Second, rigorous statistical methods—including multivariable adjustments and sensitivity analyses—were employed to control for potential confounding, thereby enhancing the credibility of our conclusions. Third, the long follow-up duration enabled robust evaluation of eGDR’s predictive value for mortality outcomes in hyperlipidemic populations.

Nonetheless, certain limitations should be acknowledged. First, lifestyle and comorbidity data were primarily self-reported, introducing potential recall bias. Second, Although the CHARLS cohort provided valuable insights into eGDR and mortality, we acknowledge that the number of mortality events in this cohort was limited. Despite a significant p for trend, the limited mortality events in CHARLS may have influenced the robustness of the results, and this should be considered when interpreting the findings. Finally, differences in life expectancy and healthcare systems between the U.S. and China may limit the generalizability of some findings across populations and should be interpreted cautiously.

Conclusions

These findings highlight the potential clinical utility of eGDR as an assessment tool for insulin resistance in predicting the risk and prognosis of hyperlipidemia. eGDR levels were significantly inversely associated with the prevalence of hyperlipidemia, as well as with all-cause and cardiovascular mortality. However, given the observational nature of this study, causal relationships and its ability to predict future risk of developing hyperlipidemia cannot be established. Further studies are needed to explore its predictive value and potential clinical applications.

Abbreviations

BMI

Body Mass Index

CHARLS

China Health and Retirement Longitudinal Study

cVA

Cardiovascular disease

eGDR

Estimated glucose disposal rate

eGFR

Estimated Glomerular Filtration Rate

FAA

Free Fatty Acids

HDL-C

High-Density Lipoprotein Cholesterol

HIEC

Hyperinsulinemic-Euglycemic Clamp

HOMA-IR

Homeostasis Model Assessment of Insulin Resistance

IR

Insulin resistance

LDL-C

Low-density lipoprotein cholesterol

HOMA-IR

Homeostasis Model Assessment of Insulin Resistance

NHANES

National Health and Nutrition Examination Survey

PIR

Family income-to-poverty ratio

VLDL-C

Very low-density lipoprotein cholesterol

Authors’ contributions

Shouxin Wei: Conceptualization, Data Curation, Formal Analysis, Software, Writing – Original Draft. Sijia Yu: Conceptualization, Methodology, Writing – Review & Editing. Chuan Qian: Resources, Validation, Writing – review & editing.Bo Chen: Formal Analysis, Conceptualization, Writing – Review & Editing.Zhengwen Xu: Data Curation, Visualization, Writing – Review & Editing.Yindong Jia: Project administration, Writing – Review & Editing.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

The dataset(s) supporting the conclusions of this article is available in the NHANES repository (https://wwwn.cdc.gov/nchs/nhanes/default.aspx, accessed 05/29/2025). The CHARLS dataset is publicly available at the official website of the China Health and Retirement Longitudinal Study (https://charls.pku.edu.cn/en/, accessed 05/29/2025). Statistical analysis code will be made available upon request. Raw data supporting the conclusions of this paper are available from the corresponding author upon request.

Declarations

Ethics approval and consent to participate

The ethical review board of the National Center for Health Statistics granted approval to the NHANES protocols, CHARLS was approved by the Ethical Review Committee of Peking University Health Science Center and ethical approval was exempted in this study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.